16.2s

python 2.6
25s

ruby crystalized [1]
4.8s
Enjoy.
-r
[1] http://github.com/rdp/crystalizer

fib :: Int -> Int
fib i | (i == 0) || (i == 1) = i
| otherwise = fib (i-1) + fib (i-2)

fibSeq :: [(Int, Int)]
fibSeq = [ (i, fib i) | i <- [0..36]]

Slim:Desktop jorrit$ ./fib
[(0,0),(1,1),(2,1),(3,2),(4,3),(5,5),(6,8),(7,13),(8,21),(9,34),(10,55),(11,89),(12,144),(13,233),(14,377),(15,610),(16,987),(17,1597),(18,2584),(19,4181),(20,6765),(21,10946),(22,17711),(23,28657),(24,46368),(25,75025),(26,121393),(27,196418),(28,317811),(29,514229),(30,832040),(31,1346269),(32,2178309),(33,3524578),(34,5702887),(35,9227465),(36,14930352)]
0.586306

[root@diego ~]# time ruby test.rb
n=0 => 0
n=1 => 1
n=2 => 1
…
n=34 => 5702887
n=35 => 9227465

real 0m15.056s
user 0m14.868s
sys 0m0.086s
[root@diego ~]# time python test.py
n=0 => 0
n=1 => 1
n=2 => 1
…
n=34 => 5702887
n=35 => 9227465

real 0m42.697s
user 0m42.107s
sys 0m0.165s
[root@diego ~]#

On a Intel Core i7 940 (2.9 Ghz) :

16s for python 2.6.2 without psyco
0.9s with psyco !!!!!

Multi-processor is the key, but there are many many implementations of stuff that will just lead you down an ultimately blind alley.

For one, I wouldn’t touch anything that furthers Microsoft’s interests.

I develop for my clients, not for sick greed.

int fib(int n) {
        if (n == 0 || n == 1) {
                return n;
        } else {
                return fib(n-1) + fib(n-2);
        }
}

int main(char* argv, int argc) {
    int i;
        for (i = 0; i  %d\n", i, fib(i));
        }
}

./test_fib
0.95s user 0.00s system 99% cpu 0.957 total

python test_fib.py
23.90s user 0.00s system 99% cpu 23.912 total

booi test_fib.boo
1.12s user 0.02s system 99% cpu 1.141 total

(now prepare to be shocked)
booc test_fib.boo; time mono test_fib.exe
0.76s user 0.00s system 99% cpu 0.768 total

test_fib.boo:

def fib(n as int) as int:
   if n == 0 or n == 1:
      return n
   else:
      return fib(n-1) + fib(n-2)

for i in range(36):
    res = fib(i)
    print "n=${i} => ${res}"

What I want to say? NEVER try to compare interpreting languages in many-function-call-tasks. Compiled strong typed languages are much better in that case, coz no runtime-detection of types needed. And compiled languages does not need to be complex and heavy to learn. In example above I took “boo” – compileable (you can use booi – iterpreter-like version as well) strong-typed (but you can use non-strong-typed constructs – slower) language with python-like syntax.

As you can see in this traditional fibonacci task boo is 40 times faster than python. And difference will be much more in big apps.

So. Right language for right job.